
ChemRxiv Preprint manuscript submitted to the Journal of Computer-Aided Molecular Design Predicting octanol/water partition coefficients using molecular simulation for the SAMPL7 challenge: comparing the use of neat and water saturated 1-octanol Spencer J. Sabatino · Andrew S. Paluch Submitted to ChemRxiv: May 3, 2021 Abstract Blind predictions of octanol/water partition coefficients at 298 K for 22 drug-like compounds were made for the SAMPL7 challenge. The octanol/water parti- tion coefficients were predicted using solvation free energies computed using molec- ular dynamics simulations, wherein we considered the use of both pure and water- saturated 1-octanol to model the octanol-rich phase. Water and 1-octanol were mod- eled using TIP4P and TrAPPE-UA, respectively, which have been shown to well reproduce the experimental mutual solubility, and the solutes were modeled using GAFF. After the close of the SAMPL7 challenge, we additionally made predictions using TIP4P/2005 water. We found that the predictions were sensitive to the choice of water force field. However, the effect of water in the octanol-rich phase was found to be even more significant and non-negligible. The effect of inclusion of water was additionally sensitive to the chemical structure of the solute. Keywords log P · partition coefficient · solvation free energy · SAMPL7 1 Introduction The partitioning of a neutral solute between two liquid phases is important for a range of industrial, environmental, and biological processes. The resulting partition coefficient is of great fundamental significance as it can both readily be measured and can be related to the underlying intermolecular interactions. During its measurement, Spencer J. Sabatino Department of Chemical, Paper, and Biomedical Engineering Miami University Oxford, Ohio 45056, USA Andrew S. Paluch Department of Chemical, Paper, and Biomedical Engineering Miami University Oxford, Ohio 45056, USA Tel.: (513) 529-0784 E-mail: [email protected] 2 Sabatino and Paluch it is desired to keep the solute concentration dilute (less than 0.01 M) in each phase so that the solute may be assumed infinitely dilute [1,2]. In doing so, the partition I=II coefficient of a solute (1) between two liquid phases at equilibrium (I and II, P1 ) may be computed as [3]: I=II c1;I log10 e h solv solv i log10 P1 (T;P) = log10 = − DG1;I (T;P) − DG1;II (T;P) c1;II RT log e = − 10 DGtran (1) RT 1;II!I where T and P correspond to the temperature and pressure, R is the molar gas con- I II stant, c1 and c1 are the molar or mass concentration (moles/volume or mass/volume) solv solv of the solute in phase I and II, respectively, DG1;I and DG1;II are the solvation free tran energy in phase I and II, respectively, and DG1;II!I is the transfer free energy from phase II to I. In assuming the solute is infinitely dilute, the partition coefficient is a measure of the relative solute-solvent interactions, while minimizing any effect the solute may have on the liquid-liquid equilibrium of the two phases. For example, the partitioning of a solute between 1-octanol and water can be used to character- ize the lipophilic/hydrophilic balance of a solute, and is an important parameter for biological systems [4,2,5]. For this reason, the prediction of octanol/water partition coefficients remains an important component of the SAMPL (Statistic Analysis of the Modeling of Proteins and Ligands) physical property prediction challenges [6]. For the case of octanol/water partition coefficients, phase I corresponds to the octanol-rich phase and the phase II corresponds to the water-rich phase. The mutual solubility of water in 1-octanol is 0.207 mole fracs, while the mutual solubility of 1-octanol in water is 0:703 × 10−4 mole fracs [7]. It is therefore reasonable to take the water-rich phase (phase II) to be pure water. On the other hand, the water in the octanol-rich phase (phase I) is not negligible. From eq. 1 this leads to: o∗=w c1;o∗ log10 e h solv solv i log10 P1 (T;P) = log10 = − DG1;o∗ (T;P) − DG1;w (T;P) c1;w RT log e = − 10 DGtran (2) RT 1;w!o∗ where the superscript “w” is used to indicate the phase taken to be pure water, and o∗=w “o*” is used to indicate water-saturated octanol. A positive value of log10 P1 is indicative of a preference of the solute for the octanol-rich phase, while a negative value is indicative of a preference for water. The greater the absolute value, the greater the preference for the respective phase. Experimentally, the measurement of octanol/water partition coefficients can be laborious. Tse and Sandler [8] therefore investigated the ability to determine oc- tanol/water partition coefficients indirectly by instead measuring values of the limit- ing activity coefficient of the solute in each phase; the log limiting activity coefficient Predicting octanol/water partition coefficients for the SAMPL7 challenge 3 and solvation free energy are directly related [3,9]. Furthermore, just as it was desir- o∗=w able to avoid the direct measurement of P1 , it was desirable to avoid measurements with water-saturated octanol and instead use only pure (neat) solvents: log e h i log Po=w (T;P) = − 10 DGsolv (T;P) − DGsolv (T;P) (3) 10 1 RT 1;o 1;w where “o” is used to indicate pure 1-octanol. For the set of organic pollutants studied, they found that there was a large difference between the limiting activity coefficient in pure water and pure 1-octanol, and in turn there was a large difference in the limiting activity coefficient in water-saturated 1-octanol and pure 1-octanol. While this led to a difference in the computed partition coefficient using neat and water-saturated 1-octanol, they did find that the partition coefficients were linearly correlated. The liquid-liquid equilibrium of water and 1-octanol has been rigorously studied by Chen and Siepmann [10] using Monte Carlo simulations with advanced sampling techniques. In both pure and water-saturated 1-octanol, microscale heterogeneities exist consisting of polar and non-polar domains. In water-saturated 1-octanol, it was shown that considerably more large hydrogen bond aggregates (i.e., polar domains) exist, where a hydrogen bond aggregate is defined as a collection of molecules where every molecule shares at least one hydrogen bond with another molecule belonging to the same aggregate. For water-saturated 1-octanol, approximately 50% of the water and 1-octanol molecules belong to clusters with aggregation numbers greater than 25; this is approximately double that in pure 1-octanol. Additionally, Chen and Siepmann [10] rigorously computed the mutual solubility of water and 1-octanol and found that the results were highly sensitive to the force fields used. They found that their combination of TIP4P water [11] and TraPPE-UA 1-octanol [12,13] resulted in an equilibrium concentration of 0.21 mole fracs of water in the octanol rich-phase in good agreement with experiment. In earlier studies, it was found that a combination of TIP4P water and OPLS-UA 1-octanol resulted in an equilibrium concentration of 0.09 mole fracs [14], and SPC water with a the modified GROMOS96 force field for 1-octanol resulted in an equilibrium concentration of 0.16 mole fracs [15]. Significant progress has been made in the use of molecular simulation (Monte Carlo and molecular dynamics) to predict octanol/water partition coefficients [16–25, 6,14]. In these efforts, the partition coefficient may be predicted with knowledge only of the structure of the solute. Moreover, these methods simultaneously may be used to understand the underlying molecular-level details. In general, the octanol/water partition coefficient is computed as the difference in solvation free energy in pure water and 1-octanol via eq. 3. While this has resulted in many accurate predictions, it physically is not in agreement with the experimental measurements. Specifically, the experimental measurements involve water-saturated 1-octanol. And following Tse solv solv and Sandler [8] we know that DG1;o∗ 6= DG1;o . Previous work has acknowledged this and attempted to model water-saturated 1-octanol using the experimental mu- tual solubility. However, it has been found that the mutual solubility predicted with common molecular models differ from experiment. If a simulation were performed at experimental conditions which corresponds to a mutual solubility greater than that predicted by the model, it would correspond to a metastable system. While the sys- 4 Sabatino and Paluch tem sizes and timescales are relatively small so that phase separation would not be observed, it would nonetheless correspond to a metastable system. The most recent SAMPL6 challenge involved the 11 molecules in fig. 1 that re- semble fragments of small molecule protein kinase inhibitors. The challenge organiz- ers encourage participants to consider the effect of water-saturation on the predicted o∗=w partition coefficients. The experimental log10 P1 values were all positive, indicat- ing a preference for the octanol-rich phase, with values ranging from 1.94–4.09. In general, it was found that the use of water-saturated 1-octanol instead of pure 1- octanol only slightly lowered the root mean squared error (RMSE) by 0.05 to 0.10 log units as compared to experiment. Methodological differences and the choice of force field were found to have a greater impact on the prediction accuracy than the composition of the 1-octanol phase [6]. For the SAMPL7 challenge participants were asked to make blind predictions of o∗=w log10 P1 for the 22 molecules in fig.
Details
-
File Typepdf
-
Upload Time-
-
Content LanguagesEnglish
-
Upload UserAnonymous/Not logged-in
-
File Pages27 Page
-
File Size-